R E S EARCH ART I C L E

ELECTR I CAL ENG INEER ING

1Center of Advanced Science and Engineering for Carbon (Case4Carbon), Departmentof Macromolecular Science and Engineering, Case Western Reserve University,Cleveland, OH 44106, USA. 2Institute of Advanced Materials for Nano-Bio Applications,School of Ophthalmology & Optometry, Wenzhou Medical University, Wenzhou, Zhejiang325027, China. 3School of Materials Science and Engineering, Georgia Institute of Tech-nology, Atlanta, GA 30332–0245, USA. 4Department of Materials Science and Engi-neering, University of North Texas, Denton, TX 76203, USA. 5Materials and ManufacturingDirectorate, Air Force Research Laboratory, Dayton, OH 45433 USA. 6Beijing Instituteof Nanoenergy and Nanosystems, Chinese Academy of Sciences, Beijing 100083,China.*Corresponding author. E-mail: zhong.wang@mse.gatech.edu (Z.L.W.); liming.dai@case.edu (L.D.)

Xue et al. Sci. Adv. 2015;1:1400198 4 September 2015

2015 © The Authors, some rights reserved;

exclusive licensee American Association for

the Advancement of Science. Distributed

under a Creative Commons Attribution

NonCommercial License 4.0 (CC BY-NC).

10.1126/sciadv.1400198

Rationally designed graphene-nanotube 3Darchitectures with a seamless nodal junction forefficient energy conversion and storage

Yuhua Xue,1,2 Yong Ding,3 Jianbing Niu,4 Zhenhai Xia,4 Ajit Roy,5 Hao Chen,2 Jia Qu,2

Zhong Lin Wang,3,6* Liming Dai1,2*

http://advancesD

ownloaded from

One-dimensional (1D) carbon nanotubes (CNTs) and 2D single-atomic layer graphene have superior thermal, electrical,andmechanical properties. However, these nanomaterials exhibit poor out-of-planeproperties due to theweak vanderWaals interaction in the transverse direction between graphitic layers. Recent theoretical studies indicate that rationallydesigned 3D architectures could have desirable out-of-plane propertieswhilemaintaining in-plane properties by grow-ing CNTs and graphene into 3D architectures with a seamless nodal junction. However, the experimental realization ofseamlessly-bonded architectures remains a challenge. We developed a strategy of creating 3D graphene-CNT hollowfibers with radially aligned CNTs (RACNTs) seamlessly sheathed by a cylindrical graphene layer through a one-stepchemical vapor deposition using an anodized aluminum wire template. By controlling the aluminum wire diameterand anodization time, the length of the RACNTs and diameter of the graphene hollow fiber can be tuned, enablingefficient energy conversion and storage. These fibers, with a controllable surface area, meso-/micropores, and supe-rior electrical properties, are excellent electrode materials for all-solid-state wire-shaped supercapacitors with poly(vinyl alcohol)/H2SO4 as the electrolyte and binder, exhibiting a surface-specific capacitance of 89.4 mF/cm2 andlength-specific capacitance up to 23.9 mF/cm, — one to four times the corresponding record-high capacitiesreported for other fiber-like supercapacitors. Dye-sensitized solar cells, fabricated using the fiber as a counterelectrode, showed a power conversion efficiency of 6.8% and outperformed their counterparts with an expensivePt wire counter electrode by a factor of 2.5. These novel fiber-shaped graphene-RACNT energy conversion andstorage devices are so flexible they can be woven into fabrics as power sources.

.sc

on O

ctober 2, 2015iencem

ag.org/

INTRODUCTION

Carbon nanomaterials, including one-dimensional (1D) carbon nano-tubes (CNTs) and 2D single-atomic layer graphene, have been demon-strated to show superior thermal, electrical, and mechanical properties.Because of the presence of strong covalent bonding in the carbonplane and the much weaker van der Waals interaction in the trans-verse direction between the layers, these 1D and 2D nanomaterialsexhibit strong direction-dependent thermal and electrical transportproperties with extremely low out-of-plane conductivities due tothe weak interlayer van der Walls interaction. However, many practi-cal applications require high through-thickness thermal and electricalconductivities. Recent theoretical studies (1, 2) have indicated that 3Dcarbon architectures, particularly a 3D pillared structure, consisting ofparallel graphene layers supported by vertically aligned CNTs (VACNTs)in between, have desirable out-of-plane transport and mechanicalproperties while maintaining the excellent properties of their buildingblocks. A few groups have tried to prepare multilayered 3D graphene-CNT pillared structures (3–7). However, most of the reported prepa-

ration methods involved multiple steps and/or chemical vapor deposition(CVD) with metal nanoparticle catalysts, which could cause detrimentaleffects on the interfacial mechanical and transport properties. So far,3D graphene-CNT network with a covalently bonded pure carbon-carbon junction at the graphene-CNT interface has rarely been exper-imentally demonstrated. To ensure that the 3D graphene-CNT structureshave efficient thermal/electrical transport characteristics in all direc-tions, it is necessary to create a seamless pure C-C nodal junction be-tween the constituent graphene sheets and CNTs. In this study, wedeveloped a novel one-step CVD method without an additional metalnanoparticle catalyst to directly grow hollow fibers made of radiallyaligned CNT (RACNT) arrays bounded by cylindrical graphene layerswith covalently bonded seamless pure C-C nodal junctions betweenthe graphene and RACNTs (Figs. 1D and 2). Different from othertypes of carbon fibers [for example, CNT fibers or graphene fibers(8–13)], these novel 3D graphene-RACNT fibers have rationally de-signed 3D micro-/mesoporous architectures with a large surface area[Brunauer-Emmett-Teller (BET) surface area, 526.91 m2/g] and mini-mized interfacial electrical/thermal resistances. Furthermore, the 3Dgraphene-RACNT structure acts as an ideal electrode material forsuper-efficient energy storage, with an area capacitance and a lengthcapacitance (capacitance per unit length) as high as 89.4 mF/cm2

and 23.9 mF/cm, respectively, both of which are one to four timesthe corresponding record-high capacities reported for other fiber-likesupercapacitors (table S1). Moreover, dye-sensitized solar cells (DSSCs)with the 3D graphene-RACNT fiber as a counter electrode to replacePt showed a power conversion efficiency up to 6.8%, which outperformedthe DSSC using CNT fiber, graphene fiber, and even Pt wire as a counterelectrode (table S2), indicating synergistic effects of the 3D architecture.

1 of 9

R E S EARCH ART I C L E

on October 2, 2015

http://advances.sciencemag.org/

Dow

nloaded from

RESULTS AND DISCUSSION

The synthetic route to the 3D graphene-RACNT hollow fibers is shownin Fig. 1 (A to C). To start with, an aluminum wire (purity >99.99%,Alfa Aesar) was anodized in 0.3 M oxalic acid solution at 40 V and 3°Cto convert the outside surface of an aluminum wire into an anodizedaluminum oxide (AAO) shell (step 1, Fig. 1, A and B). The thickness ofthe resulting AAO shell can be regulated by changing the anodizing

Xue et al. Sci. Adv. 2015;1:1400198 4 September 2015

time. Thus, a prepared wire with aluminum core and AAO shell(AAO wire) was then used as a template for a single-step CVD dep-osition without additional metal nanoparticle catalyst to produce the3D graphene fiber sheathed with RACNTs (step 2, Fig. 1, B and C).Upon CVD deposition, the color of the template wire changed fromwhite to black (fig. S1), indicating the formation of a graphitic carbonlayer over the template surface, including the inner holes of the AAO shell,to produce the 3D RACNTs sheathed with graphene, as schematically

Fig. 1. Schematic diagrams showing the synthesis and microstructures of a 3D graphene-RACNT fiber. (A) Aluminum wire. (B) Surface anodizedaluminum wire (AAO wire). (C) 3D graphene-RACNT structure on the AAO wire. (D) Schematic representation of the pure 3D graphene-RACNT structure.

(E to G) Top view SEM images of the 3D graphene-RACNT fiber at different magnifications. (I to K) SEM images of the cross-section of the 3D graphene-RACNT fiber. (H and L) AFM images of the 3D graphene-RACNT fiber. (M to P) SEM image (M) and corresponding EDX elemental mapping of (N)aluminum, (O) oxygen, and (P) carbon from the 3D graphene-RACNT fiber.

2 of 9

R E S EARCH ART I C L E

on October 2, 2015

http://advances.sciencemag.org/

Dow

nloaded from

shown in Fig. 1D. The length of the 3D hierarchically structuredgraphene-RACNT fiber thus produced is, in principle, only limitedby the length of the aluminum wire as the starting material, whichcan be commercially produced in a continuous extrusion process toallow for a large-scale production of the 3D graphene-RACNT fibers.Because of the one-step CVD growth, covalently bonded seamless C-Cnodal junctions between the graphene and the RACNTs were realizedto ensure efficient thermal/electrical transport characteristics for the3D graphene-RACNT structures. The open ends of the RACNTson the fiber surface could not only provide a large surface area butalso allow for efficient diffusions of electrolytes to and from the con-stituent nanotubes. These unique structural features make our graphene-

Xue et al. Sci. Adv. 2015;1:1400198 4 September 2015

RACNT fibers attractive for a wide range of potential applications,ranging from wearable electronics to energy conversion and storagedevices, as exemplified by ultraperformance supercapacitors andDSSCs developed in this study.

The structures of the 3D graphene-RACNT fibers were systemat-ically characterized using microscopy techniques. Figure 1E reproducesa typical low-magnification scanning electron microscopy (SEM) im-age for the as-prepared 3D graphene-RACNT fiber, which shows ahomogenous diameter of ~100 mm along the fiber length. Underhigher magnifications, opened tips of the “buried” RACNTs are evident(Fig. 1, F and G). The hole diameter seen in the SEM image (Fig. 1G) isin good agreement with that seen in an atomic force microscopy

Fig. 2. Microscopy characterization of the 3D graphene-RACNT structures. (A and B) Graphene sheet connecting to the open tips of RACNTs. (C)Closed end of RACNTs. (D) Schematic representation of the 3D graphene-RACNT network; inset shows the energy-minimized structure from MD sim-

ulations (Supplementary Materials). (E) Broken graphene sheet from the 3D graphene-RACNT network. (F) TEM image of the side view of the 3Dgraphene-RACNT around the graphene-nanotube interface. (G and H) Cross-section view of the constituent RACNTs within the 3D graphene-RACNTstructure (G) and corresponding carbon mapping (H). (I) Circle-shaped 3D graphene-RACNT fiber. (J) Piece of weaved graphene-RACNT fibers. (K) SEMimage of a knot of the graphene-RACNT fiber. (L) Photograph of a 2-m-long graphene-RACNT fiber rolled on a long stick. [The diameter of the graphene-RACNT fibers in (I) to (L) is 100 mm.]

3 of 9

R E S EARCH ART I C L E

on October 2, 2015

http://advances.sciencemag.org/

Dow

nloaded from

(AFM) image shown in Fig. 1H and can be estimated from the AFMimage to be about 50 nm, which is in consistent with the correspond-ing hole diameter of the AAO template (Fig. 1L). Figure 1 (I to K) showsthe SEM images of the cross-section of the 3D graphene-RACNT fiberunder different magnifications.

As can be seen in Fig. 1I, the unanodized aluminum core is sur-rounded by the ~6 mm-thick 3D graphene-RACNT shell, which is thesame as the RACNT length and depends mainly on the anodizedtime. The SEM images of the cross-section of the shell under highermagnifications show the presence of well-aligned CNTs (Fig. 1, J andK). The corresponding energy-dispersive x-ray spectroscopic (EDX)elemental mapping shows the presence of carbon on the fiber surfaceand through-thickness within the shell (Fig. 1P), indicating the forma-tion of the 3D graphene-RACNT fiber (cf. Fig. 1D). Figure 1 (N to P)further shows that the center of the fiber is only composed of Al (Fig.1N), whereas the shell is composed of Al (Fig. 1N) and O (Fig. 1O)in addition to C (Fig. 1P). Thus, the EDX mapping reveals that carbonhas been successfully deposited into the AAO holes within the shellof the template fiber and its surface to produce the 3D seamlesslybonded graphene-RACNT structure, as schematically shown in Fig. 1D.

By removing the aluminum and aluminum oxide in the 1 M KOHsolution, we obtained the neat hollow 3D graphene-RACNT fiber withbig graphene tube sheathing RACNT arrays along its inner surface(fig. S2). Figure 2 (A to H) shows the transmission electron micros-copy (TEM) images of the 3D graphene-RACNT fiber after removalof the aluminum and aluminum oxide template. As can be seen inFig. 2A, the RACNTs were connected to a graphene sheet of homo-geneously distributed open holes (Fig. 2B and movie S1), consistentwith the SEM (Fig. 1G) and AFM (Fig. 1, H and L) images shownin Fig. 1. The graphene edge can be observed in Fig. 2B, as indi-cated by the green arrow, and also in Fig. 2E for a small piece of theedge broken off from the graphene-RACNT sample. More TEM imagesfrom the end and side wall of the 3D graphene-RACNT fiber can befound in figs. S3 and S4, respectively. The nanotube bundles seen inFig. 2A and fig. S4A from the TEM images for broken pieces werecaused by the surface energy–induced bundling (14) during the templateremoval, whereas the integrity of the 3D graphene-RACNT structurecould be largely retained by the sheathed graphene “tube” intimatelyconnected with the open end of the constituent RACNTs, as shown infigs. S2D and S4D. The integrity of the 3D graphene-RACNT structurecan be further observed by dynamically rotating the TEM samples from−19° to 30° (fig. S5 and movie S2). The TEM images for the other endof the RACNT array are given in Fig. 2C, showing that the nanotubeend tips away from the graphene sheet in the 3D graphene-RACNTfiber were all closed (Fig. 2C). These TEM images are consistent withthe 3D graphene-RACNT network schematically drawn in Fig. 2D,which acts as a building block to form the 3D graphene-RACNT hollowfiber, as schematically shown in Fig. 1D and supported by the SEMand AFM images in Fig. 1. Also included in the inset of Fig. 2D is theenergy-minimized graphene-CNT junction from the molecular dy-namic (MD) simulations (see below and the Supplementary Mate-rials). As expected, Fig. 2D shows that the open end for each of theconstituent RACNTs was seamlessly connected to the graphene sheet.The TEM image of the cross-section view of the 3D graphene-RACNT structure given in Fig. 2F counterintuitively shows a connect-ing angle about 135° (instead of 90°) at the graphene-RACNT interface.Figure 2 (G and H) reproduces the TEM images and the electronenergy loss spectrum elemental mapping of carbon from the same part

Xue et al. Sci. Adv. 2015;1:1400198 4 September 2015

of a 3D graphene-RACNT network, respectively. The seamless carbon-carbon nodal junction was revealed by using electron tomography re-corded in TEM mode. By tilting the junction structure continuouslyfrom −70° to 70° at 1° per step, a total of 141 TEM images were re-corded from the junction, by which the 3D structure was reconstructed.A movie made from those consequent TEM images can reveal theseamless 3D junction structure between the CNTs and the graphenesheet (movies S1 and S2). The observed ~135° connecting angle is fur-ther confirmed by energy minimization and MD simulations describedin the Supplementary Materials. Figure S6 reproduces a TEM image of anindividual CNT from the graphene-RACNT fiber, which shows a rea-sonably well-graphitized nanotube structure with an outer diameter of~50 nm and about 15 concentric carbon shells.

Figure S7 shows a typical Raman spectrum of the as-prepared 3Dgraphene-RACNT fiber. As can be seen, the D band at 1350 cm−1 isrelatively weak with respect to the G band at 1595 cm−1, indicating areasonably high graphitization degree for the whole fiber sample. Theweak 2D band over 2700 cm−1 suggests the presence of graphene,although RACNT signals are dominant, as also indicated by Figs. 1Dand 2F. The x-ray photoelectron spectroscopy (XPS) spectra of the 3Dgraphene-RACNT fiber are given in fig. S8, which also shows a maincarbon peak, along with small Al and O peaks from the AAO template.The high-resolution XPS C1s spectrum shows a sharp peak at 284.5 eV,which is attributed to the sp2 C-C bond from graphitic carbons (15, 16).

Although the as-prepared AAO wire was electrically insulating, itbecame highly conductive (16 ohms/cm) after being deposited with the3D graphene-RACNT structure (fig. S9). The as-prepared 3D graphene-RACNT fiber was also sufficiently flexible to be bent from a straightline into a circle (Fig. 2I) and to be made into a piece of weave withpotentials for electronic textile applications (Fig. 2J). Similar to otherCNT or graphene fibers (8–13), our 3D graphene-RACNT fibers canalso be knotted (Fig. 2K) or rolled along a long stick (Fig. 2L).

The as-prepared 3D graphene-RACNT fibers with high conductiv-ity, good mechanical property, and pinhole-free coverage on the AAOtemplate were found to be attractive electrodes for fiber-like superca-pacitors. The electrochemical performance of the 3D graphene-RACNTfiber was first evaluated in a three-electrode cell in 1 MH2SO4 (fig. S10).Several different graphene-RACNT fiber electrodes with differentshell thicknesses (fig. S11) were also investigated. Figure S12 showsthat their electroactive surface areas, and hence supercapacitances(fig. S13), increased with increasing anodization time (fig. S12 andtable S3). To further enlarge the electroactive surface area, anothertwo large-diameter aluminum wires (380 and 810 mm) were selected(figs. S14 and S15). The cyclic voltammogram (CV) curves of thesethree fibers are shown in fig. S16A, which reveals a much higher cur-rent density for fiber-810 than for fiber-380 and fiber-100. Never-theless, fiber-100 has the lowest resistance of about 1.4 ohms (fig.S16B and table S4).

Figure 3 (A and B) shows the SEM images of a solid-state wire-likesupercapacitor based on two twisted 3D graphene-RACNT fiberelectrodes sandwiched with poly(vinyl alcohol) (PVA)/H2SO4 gel asthe solid electrolyte and binder. The CVs measured from the solid-statewire supercapacitor at different scanning rates from 100 to 500 mV/sall show a regular rectangular shape (Fig. 3C), indicating an ideal ca-pacitive behavior with a high-rate capability. The CV curve is stillsomewhat rectangular in shape even at the ultrahigh scan rate up to50 V/s (fig. S17A) because of the ultralow resistance (fig. S16B andtable S4). Furthermore, the newly developed wire supercapacitor could

4 of 9

R E S EARCH ART I C L E

on October 2, 2015

http://advances.sciencemag.org/

Dow

nloaded from

Fig. 3. Performance of the solid wire supercapacitors of 3D graphene-CNT fiber for energy storage. (A and B) SEM images of the top and cross-section of the solid-state wire supercapacitor based on the 3D graphene-RACNT fiber electrodes (the diameter of the fiber is 100 mm). (C) CV curves of the

solid-state wire supercapacitor based on the 3D graphene-RACNT wire electrodes at scanning rates from 100 to 500 mV/s. (D) Galvanostatic charge anddischarge curves of the 3D graphene-RACNT wire capacitor at different currents. (E) Surface-specific capacitance of the 3D graphene-RACNT wire electrodecalculated from the galvanostatic charge/discharge curves. (F to L) Integrated solid wire supercapacitors either in series or in parallel (the diameter of thefibers used in integrated devices is 380 mm). (F) Schematic representation of an integrated supercapacitor in series from three graphene-RACNT wiresupercapacitors. (G and H) Galvanostatic charge-discharge curves and CV curves, respectively, for the series integrated graphene-RACNT wire supercapacitorand a single wire supercapacitor. (I) Use of an integrated supercapacitor to light up a commercial LED. (J) Schematic representation of an integrated super-capacitor in parallel from three graphene-RACNT wire supercapacitors. (K and L) Galvanostatic charge-discharge curves and CV curves, respectively, for theparallel integrated graphene-RACNT wire supercapacitor and a single wire supercapacitor.

Xue et al. Sci. Adv. 2015;1:1400198 4 September 2015 5 of 9

R E S EARCH ART I C L E

http:D

ownloaded from

be charged and discharged at either high (fig. S17B) or low current den-sities (Fig. 3D). The galvanostatic charge/discharge curves measured athigh current densities from 424 to 2120 mA/cm2 are nearly symmetricstraight lines (fig. S17B), indicating an ideal double-layer superca-pacitor behavior. Even at a very low current density of 2.12 mA/cm2, thegalvanostatic charge/discharge curve shows a triangular shape (Fig.3D). From the galvanostatic charge/discharge curve, a specific ca-pacitance of ~3.33 mF/cm2 was calculated at the discharge currentof 2.12 mA/cm2 for the solid-state wire supercapacitor based on the3D graphene-RACNT fiber electrodes. The specific capacitancewas found to decrease with increasing current density (Fig. 3E). Byusing the 810-mm-thick aluminum wire after a prolonged anodization(120 hours) as the electrodes, we can further increase capacitance of thesolid-state 3D graphene-RACNT fiber supercapacitor. In fig. S18, arecord-high surface-specific capacitance up to 89.4 mF/cm2 was ob-tained for the 3D graphene-RACNT fiber (810 mm), which is much high-er than the capacitance of wire supercapacitors made from ZnOnanorods (2 mF/cm2) (17), CNT fibers (0.6 mF/cm2) (18), and graphenefibers (1.7 mF/cm2) (19), and even higher than that of the PEDOT[poly(3,4-ethylenedioxythiophene)]–coated CNT fibers (73 mF/cm2)(20). The length-specific capacitance of the 3D graphene-RACNT wire(810 mm) supercapacitor is as high as 23.9 mF/cm, which is almost1000 times that of the CNT fibers (0.024 mF/cm) (18) and graphene

Xue et al. Sci. Adv. 2015;1:1400198 4 September 2015

fibers (0.02 mF/cm) (19) and even 50 times higher than that of thePEDOT-coated CNT capacitors (0.46 mF/cm) (20). Although somemoreadvanced capacitors have been reported in recent years (21–23), our 3Dgraphene-RACNT wire capacitor is still among the best.

The newly developed solid-state wire supercapacitor based on thegraphene-RACNT electrodes was further subjected to a long-term sta-bility test. As can be seen in fig. S19A, there was no obvious changeeven after about 10,000 consecutive charge/discharge cycles, indicatingexcellent long-term device stability. Moreover, the 3D graphene-RACNTwire supercapacitor also showed high flexibility, wherein it could beeasily rolled over a glass tube without any noticeable effect on the de-vice performance (fig. S20). Figure S19B shows bending cycle stabilitymeasured for the 3D graphene-RACNT supercapacitor over 100 timesat bending angles up to 90°, indicating very good stability with about90% capacitance retention over bending for 100 times.

Having demonstrated superb performance for the solid-state 3Dgraphene-RACNT wire supercapacitor, we have further integrated in-dividual 3D graphene-RACNT wire supercapacitors into assemblieseither in series or in parallel to meet specific energy needs for diversepractical applications. Figure 3 (F and I) shows three wire supercapa-citors connected in series. As expected, the charge/discharge voltagewindow of the three graphene-RACNT wire supercapacitors connectedin series is about three times of that for a single wire supercapacitor

on October 2, 2015

//advances.sciencemag.org/

Fig. 4. 3D graphene-RACNT fiber as the counter electrode for wire-shaped DSSCs. (A) Schematic representation of wire-shaped DSSC using 3Dgraphene-CNT fiber as the counter electrode and the TiO nanotube fiber as the photo anode. (B) Schematic drawing for the cross-section view of

2

the wire-shaped DSSC. (C) SEM image of the cross-section view of the wire-shaped DSSC. (D) Photograph of the wire-shaped DSSC sealed in a glasscapillary tube. (E) Flexible wire-shaped DSSC sealed in a transparent FET plastic tube. (F) Knot made from the flexible DSSC. (G) Current density–voltagecharacteristics of DSSCs with graphene wire, CNT wire, Pt wire, or 3D graphene-RACNT fiber as the counter electrode. (H) Nyquist plots of the 3D graphene-CNT fiber and the Pt wire measured in the I−/I3

− electrolyte (10 mM LiI + 1 mM I2 + 0.1 M LiClO4 + acetonitrile). (I) Current density–voltage characteristics ofDSSCs with the 3D graphene-RACNT wire counter electrode before and after bending.

6 of 9

R E S EARCH ART I C L E

on October 2, 2015

http://advances.sciencemag.org/

Dow

nloaded from

(Fig. 3, G and H) with an almost identical discharge time. The in-tegrated devices can be used as a power source, as exemplified in Fig.3I, to light up a commercial LED (Red LED, 1.9 to 2.0 V, from Micro-tivity). As can be seen in Fig. 3 (J to L), the output current density andthe discharge time for three graphene-RACNT wire supercapacitorsconnected in a parallel configuration are about three times of thosefor a single wire supercapacitor. These results show the scalability ofthe 3D graphene-RACNT wire supercapacitors. To further demon-strate the potential applications for the newly developed all-solid-stategraphene-RACNT wire supercapacitors as efficient energy storagecomponents for wearable devices, we weaved the 3D graphene-RACNTwire supercapacitors into fabrics (fig. S21A). These wire capacitorsweaved into fabrics could be used as flexible power sources withoutdetrimental effect on their performance.

The 3D graphene-RACNT fiber could also be used as an efficientcounter electrode in DSSCs to replace the expensive Pt while still re-taining and even exceeding the high performance of the Pt-based DSSCs.In Fig. 4 (A and B), the wire-shaped DSSC was fabricated by twining atitanium wire sheathed with vertically aligned titanium oxide (TiO2) na-notubes (fig. S22) as the photoanode with the graphene-RACNT counterelectrode (Fig. 4C), which was sealed in a capillary tube containing theI3−/I− electrolyte solution (Fig. 4D) (24, 25). The performance of thewire-shaped DSSC thus prepared was evaluated under standard illu-mination (AM 1.5, 100 mW/cm2). Figure 4G shows the current density–voltage (J-V) characteristics with the corresponding numerical datalisted in table S2. As can be seen, the wire-shaped DSSC with a 3Dgraphene-CNT fiber counter electrode exhibited a short-circuit cur-rent (Jsc) of 14.50 mA/cm2, an open-circuit voltage (Voc) of 0.70 V,and a fill factor (FF) of 0.67. The overall power conversion efficiencyis as high as 6.80%, which is much higher than that of its counterpartswith the Pt wire (2.74%), graphene wire (0.63%), or CNT wire (1.14%)as the counter electrode (Fig. 4G and table S2). The series resistanceand charge transfer resistance of the 3D graphene-RACNT fiber aremuch smaller than those of the Pt wire (Fig. 4H). Compared with apure graphene electrode, the 3D graphene-RACNT fiber electrode hasrationally designed 3D micro-/mesoporous architectures with a largesurface area and a seamless nodal junction at the graphene and RACNTinterface. Thus, the 3D graphene-RACNT fiber electrode providesenormous graphitic surface area to speed the I3

−/I− reaction, the 3Dmicro-/mesoporous architectures to facilitate the reactants and elec-trolyte diffusion, and the seamless nodal junction to minimize the in-terfacial electrical/thermal resistances, which significantly enhanced thedevice performance. Flexible wire-shaped DSSCs can also be preparedby sealing the electrodes in a transparent fluorinated ethylene propylene(FET) tube (Fig. 4E), which is sufficiently flexible to be made into aknot (Fig. 4F). Bending the flexible DSSC shown in fig. S23 did notcause any obvious change in its performance (Fig. 4I and table S5).Because of its versatile and scalable nature, the methodologies devel-oped in this study could be applicable to the scalable production ofother 3D hierarchically structured multifunctional materials for a widerange of applications.

MATERIALS AND METHODS

Preparation of the AAO wireThree different aluminum wires (purity > 99.99%) with diameters of13, 100, 380, and 810 mm were purchased from Alfa Aesar Company.

Xue et al. Sci. Adv. 2015;1:1400198 4 September 2015

Before use, the aluminum wire was washed in acetone for 10 min andwas then anodized by a two-step electrochemical process, as previous-ly described (26, 27). In a typical experiment, aluminum wires wereanodized in 0.3 M oxalic acid solution at 3°C and 40 V. The firstanodizing step took 1 hour. The aluminum oxide layer was removedin the mixture solution of 6 wt % phosphoric acid and 1.8 wt % chro-mic acid at 60°C for 4 hours. The second anodizing step lasted for 2 to120 hours depending on the desirable thicknesses for the AAO shell.The pores in the AAO shell were finally widened in 0.1 M phosphoricacid solution for 40 min.

Synthesis of the 3D graphene-RACNT fiber on the AAO wireThe 3D graphene-RACNT fiber was grown on the AAO wire by usinga one-step CVD process. Typically, the AAO wire was put inside a quartztube furnace. The furnace was heated up to 620°C with argon gas con-tinuously flowing at a rate of 200 SCCM (standard cubic centimeterperminute). At 620°C, acetylene gas (20 SCCM) and hydrogen (10 SCCM)were introduced to the furnace. The sample was kept in the furnaceunder these conditions for 55 min to achieve the one-step growth ofthe 3D graphene-RACNT fiber on the AAO wire. Upon completionof the growth process, the 3D graphene-RACNT fiber was rapidly movedaway from the furnace center (620°C), and the quartz tube was air-cooleddown to room temperature. Thereafter, the 3D graphene-RACNT fi-ber was taken out from the quartz tube for characterization and wasdirectly used as the wire electrode without any further treatment un-less otherwise stated.

Fabrication of the solid-state wire supercapacitorThe PVA/H2SO4 gel was used as the solid electrolyte, which wasprepared as follows: 2.2 g of PVA (weight-average molecular weight,85,000 to 124,000; Sigma-Aldrich) powder was added into 20 ml ofwater, followed by heating at 90°C under vigorous stirring, and then2.2 g of H2SO4 was added after the PVA solution became clear.

For supercapacitor construction, the 3D graphene-RACNT fiberswere precoated with a layer of PVA/H2SO4 gel and then dried at roomtemperature for 1 hour. The flexible solid-state wire supercapacitorwas then fabricated by intertwisting two of the 3D graphene-RACNTfiber electrodes together.

The surface-specific capacitance for the solid-state wire capacitancewas calculated from the galvanostatic charge and discharge curves byusing the following equation:

Cs ¼ 4IDt=ðSV Þ;where I is the applied current, Dt is the discharge time, S is the surfacearea of the graphene-RACNT wire electrodes, and V is the potentialrange.

The corresponding length-specific capacitance was calculated fromthe galvanostatic charge and discharge curves using the followingequation:

CL ¼ 4IDt=ðLV Þ;where L is the length of the graphene-CNT wire electrodes.

The surface- and length-specific capacitances for the three-electrodesupercapacitor in H2SO4 were calculated from the galvanostatic chargeand discharge curves using the following equations:

Cs ¼ IDt=ðSV Þ and CL ¼ IDt=ðLV Þ:

7 of 9

R E S EARCH ART I C L E

http://advances.sciencemD

ownloaded from

Fabrication of the wire DSSCThe aligned TiO2 nanotube/Ti wire was prepared by electrochemicalanodizing of titanium wire (diameter of 127 mm, purity of 99.99%;Alfa Aesar) in 0.3 wt % NH4F/ethylene glycol solution containing 2 v %H2O at voltages of 60 V for 5.5 hours (28). After being washed withdeionized water, the resulting TiO2 nanotube/Ti wire was then an-nealed at 500°C in air for 1 hour. The annealed wire was further treatedwith 40 mM TiCl4 solution at 70°C for 30 min, followed by annealing at450°C in air for 30 min. The TiO2 nanotube/Ti wire was immersed ina mixture solvent of acetonitrile and isobutanol (1:1, v/v) solutioncontaining 0.5 mM N719 dye (Solaronix) for 24 hours to form thesensitized wire photoanode. The wire-shaped DSSC cell was fabricatedby packaging the aligned TiO2 nanotube/Ti wire photoanode and the3D graphene-RACNT fiber counter electrode in a capillary tube or trans-parent FET tube (purchased from Amazon) containing the electrolyteof 0.6 M dimethyl-3-propylimidazolium iodide, 0.04 M I2, 0.02 M LiI,0.1 M guanidine thiocyanate, and 0.5 M 4-terbutylpyridine in a mixtureof acetonitrile and valeronitrile. To fabricate the reference DSSCs, thegraphene wire was grown on the copper wire under 1000°C with meth-ane as the carbon source and used as the counter electrode while theCNT fiber was obtained by dry spinning from VACNTs.

CharacterizationThe SEM images were recorded on a Hitachi SU8010 cold field emis-sion SEM. SEM mapping was done on a Philips XL30 environmentalscanning electron microscope. TEM was carried out using an FEITecnai F30 TEM working at 300 kV. The AFM images were ob-tained on an Agilent Technologies 5500 Scanning Probe Microscope.The XPS measurements were performed on a PHI 5000 VersaProbe.The Raman spectra were collected using a Raman spectrometer(Renishaw) with a 514-nm laser.

on October 2, 2015

ag.org/

SUPPLEMENTARY MATERIALSSupplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/1/8/e1400198/DC1Fig. S1. Photographs of aluminum wire, AAO wire, and 3D graphene-RACNT deposited on theAAO wire.Fig. S2. SEM images of the cross-section view of the 3D graphene-RACNT fiber after theremoval of the aluminum and AAO template under different magnifications.Fig. S3. TEM image of the edge of the 3D graphene-RACNT fiber with different magnifications.Fig. S4. Top view TEM images of the center of the 3D graphene-RACNT fiber with differentmagnifications.Fig. S5. TEM images of the side of the 3D graphene-RACNT fiber rotated with different anglesfrom −19° to 30° around the red arrow.Fig. S6. TEM images of an individual carbon nanotube from the 3D graphene-RACNT fiberunder different magnifications.Fig. S7. Raman spectrum of the 3D graphene-RACNT fiber on the AAO wire.Fig. S8. The XPS survey spectrum of the 3D graphene-RACNT fiber, and the correspondinghigh-resolution XPS C1s spectrum.Fig. S9. The resistance of the as-prepared 3D graphene-RACNT wire.Fig. S10. CV curves and galvanostatic discharge curves of the 3D graphene-RACNT wire (0.1 mm indiameter) electrode in 1 M H2SO4 solution.Fig. S11. SEM images of the 3D graphene-RACNT wire (100 mm in diameter) with different shellthicknesses via anodizing at different times.Fig. S12. CV curves of the fiber-100mm-2hrs, fiber-100mm-4hrs, and fiber-100mm-12hrs in 5 mMK3Fe(CN)6/0.1 M KCl solution.Fig. S13. CV curves, galvanostatic charge and discharge curves, and the surface specificcapacitance of the 3D graphene-RACNT wire electrodes (fiber-100mm-2hrs, fiber-100mm-4hrs,and fiber-100mm-12hrs).Fig. S14. Photographs of the 3D graphene-RACNT fibers prepared from aluminum wire withdifferent diameters.

Xue et al. Sci. Adv. 2015;1:1400198 4 September 2015

Fig. S15. SEM images of the 3D graphene-RACNT fibers prepared from aluminum wires withdifferent diameters.Fig. S16. CV curves and Nyquist plots of the 3D graphene-RACNT fibers with a diameter.Fig. S17. CV curves of the solid-sate wire supercapacitor at high scanning rates, and galvanostaticcharge and discharge curves of the 3D graphene-VACNT wire capacitor at high current density.Fig. S18. Photograph, CV curves, galvanostatic charge and discharge curves, and the surfacespecific capacitance and length specific capacitance of the solid-state wire supercapacitorbased on the 3D graphene-RACNT fiber electrodes (810 mm in diameter).Fig. S19. The long term stability and bending stability tests for the supercapacitor based on the3D graphene-RACNT electrodes with diameter of 100 mm.Fig. S20. Photographs, CV curves, and galvanostatic charge/discharge curves of the 3Dgraphene-RACNT wire supercapacitor before and after rolling on a glass tube.Fig. S21. Photographs of graphene-RACNT wire supercapacitor weaved into a piece of fabricand rolling over a stick, and its CV curves and galvanostatic charge/discharge curves beforeand after rolling on a stick.Fig. S22. SEM images of TiO2 nanotube/Ti wire.Fig. S23. SEM images of a 3D graphene-RACNT flexible wire DSSC before and after bend.Fig. S24. Schematics of Al2O3 template with a filleted hole in the center, andMDmodel of the template.Fig. S25. Schematics of CNT-graphene junction.Fig. S26. Side view and cross section of three-layered CNT-graphene junctions with 135° filletand without fillet.Fig. S27. Normalized bending energy predicted by the analytical model, and normalizedpotential energy calculated by MD as a function of the fillet angles.TableS1. The reportedareal capacitanceand lengthcapacitanceof the fiber supercapacitor in references.Table S2. Jsc, Voc, FF, and power conversion efficiency for wire-shaped DSSCs with the 3Dgraphene-CNT fiber, and Pt wire as the counter electrode.Table S3. Electroactive surface areas of the 3D graphene-RACNT fiber electrodes.Table S4. Series resistance of the 3D graphene-RACNT fiber electrodes.Table S5. Jsc, Voc, FF, and power conversion efficiency for wire-shaped DSSCs with the 3Dgraphene-CNT fiber before and after bend.Movie S1. A movie made from those consequent TEM images, showing the seamless 3Djunction structure between the CNTs and the graphene sheet.Movie S2. A movie made from those consequent TEM images, revealing the aligned CNTbundles (see text) and their seamless junction with the graphene sheet.

REFERENCES AND NOTES1. V. Varshney, S. S. Patnaik, A. K. Roy, G. Froudakis, B. L. Farmer, Modeling of thermal

transport in pillared-graphene architectures. ACS Nano 4, 1153–1161 (2010).2. G. K. Dimitrakakis, E. Tylianakis, G. E. Froudakis, Pillared graphene: A new 3-D network

nanostructure for enhanced hydrogen storage. Nano Lett. 8, 3166–3170 (2008).3. Y. Zhu, L. Li, C. Zhang, G. Casillas, Z. Sun, Z. Yan, G. Ruan, Z. Peng, A.-R. Raji, C. Kittrell, R. H. Hauge,

J. M. Tour, A seamless three-dimensional carbon nanotube graphene hybrid material.Nat. Commun. 3, 1225 (2012).

4. F. Du, D. Yu, L. Dai, S. Ganguli, V. Varshney, A. K. Roy, Preparation of tunable 3D pillaredcarbon nanotube-graphene networks for high-performance capacitance. Chem. Mater. 23,4810–4816 (2011).

5. S. Li, Y. Luo, W. Lv, W. Yu, S. Wu, P. Hou, Q. Yang, Q. Meng, C. Liu, H.-M. Cheng, Verticallyaligned carbon nanotubes grown on graphene paper as electrodes in lithium-ion batteriesand dye-sensitized solar cells. Adv. Energy Mater. 1, 486–490 (2011).

6. R. K. Paul, M. Ghazinejad, M. Penchev, J. Lin, M. Ozkan, C. S. Ozkan, Synthesis of a pillaredgraphene nanostructure: A counterpart of three-dimensional carbon architectures. Small6, 2309–2313 (2010).

7. C. Tang, Q. Zhang, M.-Q. Zhao, G.-L. Tian, F. Wei, Resilient aligned carbon nanotube/graphenesandwiches for robust mechanical energy storage. Nano Energy 7, 161–169 (2014).

8. Y.-L. Li, I. A. Kinloch, A. H. Windle, Direct spinning of carbon nanotube fibers from chemicalvapor deposition synthesis. Science 304, 276–278 (2004).

9. K. Koziol, J. Vilatela, A. Moisala, M. Motta, P. Cunniff, M. Sennett, A. Windle, High-performancecarbon nanotube fiber. Science 318, 1892–1895 (2007).

10. N. Behabtu, C. C. Young, D. E. Tsentalovich, O. Kleinerman, X. Wang, A. W. K. Ma, E. A. Bengio,R. F. ter Waarbeek, J. J. de Jong, R. E. Hoogerwerf, S. B. Fairchild, J. B. Ferguson, B. Maruyama,J. Kono, Y. Talmon, Y. Cohen, M. J. Otto, M. Pasquali, Strong, light, multifunctional fibers ofcarbon nanotubes with ultrahigh conductivity. Science 339, 182–186 (2013).

11. Z. Xu, C. Gao, Graphene chiral liquid crystals and macroscopic assembled fibre. Nat. Commun.2, 571 (2011).

12. M. D. Lima, S. Fang, X. Lepró, C. Lewis, R. Ovalle-Robles, J. Carretero-González, E. Castillo-Martínez,M. E. Kozlov, J. Oh, N. Rawat, C. S. Haines, M. H. Haque, V. Aare, S. Stoughton, A. A. Zakhidov,R. H. Baughman, Biscrolling nanotube sheets and functional guests into yarns. Science 331,51–55 (2011).

8 of 9

R E S EARCH ART I C L E

http://adD

ownloaded from

13. H. P. Cong, X. C. Ren, P. Wang, S. H. Yu, Wet-spinning assembly of continuous, neat, andmacroscopic graphene fibers. Sci. Rep. 2, 613 (2012).

14. H. Li, C.-L. Xu, G.-Y. Zhao, H.-L. Li, Effects of annealing temperature on magnetic propertyand structure of amorphous Co49Pt51 alloy nanowire arrays prepared by direct-currentelectrodeposition. J. Phys. Chem. B 109, 3759–3763 (2005).

15. A. N. Sokolov, F. L. Yap, N. Liu, K. Kim, L. Ci, O. B. Johnson, H. Wang, M. Vosgueritchian, A. L. Koh,J. Chen, J. Park, Z. Bao, Direct growth of aligned graphitic nanoribbons from a DNA template bychemical vapour deposition. Nat. Commun. 4, 2402 (2013).

16. M. A. Fanton, J. A. Robinson, C. Puls, Y. Liu, M. J. Hollander, B. E. Weiland, M. Labelle, K. Trumbull,R. Kasarda, C. Howsare, J. Stitt, D. W. Snyder, Characterization of graphene films and tran-sistors grown on sapphire by metal-free chemical vapor deposition. ACS Nano 5, 8062–8069(2011).

17. J. Bae, M. K. Song, Y. J. Park, J. M. Kim, M. Liu, Z. L. Wang, Fiber supercapacitors made ofnanowire-fiber hybrid structures for wearable/flexible energy storage. Angew. Chem. Int. Ed.50, 1683–1687 (2011).

18. T. Chen, L. Qiu, Z. Yang, Z. Cai, J. Ren, H. Li, H. Lin, X. Sun, H. Peng, An integrated “energy wire” forboth photoelectric conversion and energy storage. Angew. Chem. Int. Ed. 51, 11977–11980 (2012).

19. Y. Meng, Y. Zhao, C. Hu, H. Cheng, Y. Hu, Z. Zhang, G. Shi, L. Qu, All-graphene core-sheathmicrofibers for all-solid-state, stretchable fibriform supercapacitors and wearableelectronic textiles. Adv. Mater. 25, 2326–2331 (2013).

20. J. A. Lee, M. K. Shin, S. H. Kim, H. U. Cho, G. M. Spinks, G. G. Wallace, M. D. Lima, X. Lepró,M. E. Kozlov, R. H. Baughman, S. J. Kim, Ultrafast charge and discharge biscrolled yarnsupercapacitors for textiles and microdevices. Nat. Commun. 4, 1970 (2013).

21. D. T. Pham, T. H. Lee, D. H. Luong, F. Yao, A. Ghosh, V. T. Le, T. H. Kim, B. Li, J. Chang, Y. H. Lee,Carbon nanotube-bridged graphene 3D building blocks for ultrafast compact supercapacitors.ACS Nano 9, 2018–2027 (2015).

22. H.-J. Peng, J.-Q. Huang, M.-Q. Zhao, Q. Zhang, X.-B. Cheng, X.-Y. Liu, W.-Z. Qian, F. Wei,Nanoarchitectured graphene/CNT@porous carbon with extraordinary electrical conductiv-ity and interconnected micro/mesopores for lithium-sulfur batteries. Adv. Func. Matter 24,2772–2781 (2014).

Xue et al. Sci. Adv. 2015;1:1400198 4 September 2015

23. D. Yu, K. Goh, H. Wang, L. Wei, W. Jiang, Q. Zhang, L. Dai, Y. Chen, Scalable synthesis ofhierarchically structured carbon nanotube–graphene fibres for capacitive energy storage.Nat. Nanotechnol. 9, 555–562 (2014).

24. Y. Xue, J. Liu, H. Chen, R. Wang, D. Li, J. Qu, L. Dai, Nitrogen-doped graphene foams asmetal-free counter electrodes in high-performance dye-sensitized solar cells. Angew.Chem. Int. Ed. 51, 12124–12127 (2012).

25. H. Wang, Y. H. Hu, Graphene as a counter electrode material for dye-sensitized solar cells.Energy Environ. Sci. 5, 8182–8188 (2012).

26. H. Masuda, K. Fukuda, Ordered metal nanohole arrays made by a two-step replication ofhoneycomb structures of anodic alumina. Science 268, 1466–1468 (1995).

27. M. J. Zheng, G. Li, X. Zhang, S. Huang, Y. Lei, L. Zhang, Fabrication and structural characteri-zation of large-scale uniform SnO2 nanowire array embedded in anodic alumina membrane.Chem. Mater. 13, 3859–3861 (2001).

28. T. Chen, L. Qui, H. G. Kia, Z. Yang, H. Peng, Designing aligned inorganic nanotubes at theelectrode interface: Towards highly efficient photovoltaic wires. Adv. Mater. 24, 4623–4628 (2012).

Funding: We are grateful for the financial support from the U.S. Department of Defense Mul-tidisciplinary University Initiative under AFOSR (FA9550-12-1-0037; J. Harrison, Program Man-ager), Case Western Reserve University–Wenzhou Medical University (CON115346), andNatural Science Foundation of China (51202167). Competing interests: The authors declarethat they have no competing interests.

Submitted 11 December 2014Accepted 16 June 2015Published 4 September 201510.1126/sciadv.1400198

Citation: Y. Xue, Y. Ding, J. Niu, Z. Xia, A. Roy, H. Chen, J. Qu, Z. L. Wang, L. Dai, Rationallydesigned graphene-nanotube 3D architectures with a seamless nodal junction for efficientenergy conversion and storage. Sci. Adv. 1, 1400198 (2015).

va

9 of 9

on October 2, 2015

nces.sciencemag.org/

doi: 10.1126/sciadv.14001982015, 1:.Sci Adv 

2015)Chen, Jia Qu, Zhong Lin Wang and Liming Dai (September 4, Yuhua Xue, Yong Ding, Jianbing Niu, Zhenhai Xia, Ajit Roy, Haostorage

anda seamless nodal junction for efficient energy conversion Rationally designed graphene-nanotube 3D architectures with

this article is published is noted on the first page. This article is publisher under a Creative Commons license. The specific license under which

article, including for commercial purposes, provided you give proper attribution.licenses, you may freely distribute, adapt, or reuse theCC BY For articles published under

. hereAssociation for the Advancement of Science (AAAS). You may request permission by clicking for non-commerical purposes. Commercial use requires prior permission from the American

licenses, you may distribute, adapt, or reuse the articleCC BY-NC For articles published under

http://advances.sciencemag.org. (This information is current as of October 2, 2015):The following resources related to this article are available online at

http://advances.sciencemag.org/content/1/8/e1400198.full.htmlonline version of this article at:

including high-resolution figures, can be found in theUpdated information and services,

http://advances.sciencemag.org/content/suppl/2015/09/01/1.8.e1400198.DC1.html can be found at: Supporting Online Material

http://advances.sciencemag.org/content/1/8/e1400198#BIBL5 of which you can be accessed free: cites 28 articles,This article

trademark of AAAS otherwise. AAAS is the exclusive licensee. The title Science Advances is a registered York Avenue NW, Washington, DC 20005. Copyright is held by the Authors unless statedpublished by the American Association for the Advancement of Science (AAAS), 1200 New

(ISSN 2375-2548) publishes new articles weekly. The journal isScience Advances

on October 2, 2015

http://advances.sciencemag.org/

Dow

nloaded from